Four account-support scenarios in one context-aware voice workflow
HelloWork receives frequent procedural questions about opening and logging in to My Page accounts. These requests are repetitive, but the correct guidance depends on who is calling, where the user is in the process and what information or error they have encountered.
The helpdesk needed an automated way to guide users through these common questions without treating every call as a simple scripted exchange. The workflow had to distinguish job seekers from employers, identify the user's intent, retrieve the relevant instructions and keep track of progress across follow-up questions.
Dyna.Ai provided Voice Agent capabilities covering conversation design, FAQ and retrieval-augmented knowledge structuring, response control, testing and agent tuning. After fine-tuning, the Voice Agent reached more than 96% accuracy in the reported evaluation. The workflow covered four priority scenarios and supported conversations averaging approximately 15 minutes.
The challenge
Repetitive My Page questions were consuming helpdesk capacity
The helpdesk handles account-opening and login questions from two distinct user groups: job seekers and employers. Although these requests are procedural, the user may need help with a different account type, a different step on the website or a different error condition.
Manual handling increases the workload on helpdesk agents and can make guidance less consistent across user types and inquiry patterns. A response that is correct for a job seeker may not be appropriate for an employer, and a user who has already completed one step should not be sent back through the entire process.
The deployment
The solution covered four high-frequency account journeys
Dyna.Ai configured a Voice Agent workflow for four priority scenarios:
The agent identifies the user's type and inquiry intent before retrieving relevant information from prepared manuals and FAQs. It then guides the user through the appropriate website steps instead of applying one generic answer to every caller.
How the workflow changed
Retrieval and response control kept procedural guidance grounded
The solution combined FAQ and RAG knowledge structuring with response control. Manuals and FAQs provided the information foundation, while conversation design determined how the agent identified the user type, selected the relevant path and explained the next step.
The Voice Agent was designed to guide users through the website rather than simply recite a knowledge-base answer. It could handle wording variations, missing information and changes in question order while maintaining the user's position in the process. Account support is a sequence of actions, so the agent needed to preserve enough context to continue the right journey after a follow-up question or error.
The results
Long conversations tested whether the flow could hold context
Calls averaged approximately 15 minutes, giving the workflow a test of long-conversation control. The agent maintained context across completed steps, user responses, errors and follow-up questions.
When an inquiry fell outside the supported scenarios, the agent guided the user to human support. The operating boundary was clear: automation handled defined procedural questions, while unsupported requests had a clear escalation path.
Testing and agent tuning were included; test volume, acceptance criteria, knowledge refresh, escalation handoff and systems integration are unspecified.
Human responsibility
More consistent account guidance across four scenarios
The helpdesk gained a structured Voice Agent workflow for four account-support scenarios across two user types. After fine-tuning, the reported accuracy exceeded 96%. The agent could identify intent, retrieve prepared guidance, maintain context through an approximately 15-minute conversation and route unsupported questions to human support.
These are implementation outcomes, not a quantified claim that the helpdesk eliminated manual work or achieved a specific service improvement. Call volume, completion, transfer and workload results remain to be confirmed.
Use operational fit as the primary angle: bounded, testable and context-aware voice workflows for job seekers and employers.
